An assessment of the internets potential in enhancing consumer relationships

School of Information Systems Faculty of Business and Law Victoria University of Technology An Assessment of the Internet’s Potential in Enhancing Consumer Relationships Submitted by: Noor Raihan Ab Hamid This thesis is presented to fulfill the requirements for the award of Doctor of Philosophy, Victoria University of Technology August 2006 Student declaration “I, Noor Raihan Ab Hamid, declare that the PhD thesis entitled ‘An assessment of the Internet’s potential in enhancing consumer relationships’ is no more than 100,000 words in length, exclusive of tables, figures, appendices, references and footnotes. This thesis contains no material that has been submitted previously, in whole or in part, for the award of any other academic degree or diploma. Except where otherwise indicated, this thesis is my own work”. Signature Date i Acknowledgment This thesis would be just impossible without the blessings from Allah The Mighty and valid support and guidance from many personalities who believe in me and my undertakings. I would like to record my warmest gratitude to my learned supervisors, Professor Michael McGrath and Dr Stephen Burgess whose sharp sense of research direction have provided invaluable feedback to improve the quality of this thesis. I would also like to thank Mr Rodney Turner for his useful comments. To my employer, Multimedia University I owe special obligation for all the facilities extended to me especially the financial support for the data collection phase as well as the final cycle of my study, paving way for the accomplishment of this thesis. My special thanks to Professor Ismail, Dean of Faculty of Management and Dr Ali Khatibi both from Multimedia University for the wisdom of their undivided support and encouragement particularly in the early phase of my study. I am greatly indebted to my colleague, Dr Norizan Kassim from whom I have learnt more about Structural Equation Modeling technique where her comments through our series of discussions had added salt to my work. I would also like to thank my research assistants, Khairatun Hisan and Johanniz Rosli who had energetically assisted me in the data collection and data entry phase. I am truly grateful to my parents for their everlasting inspiration and giving me all the opportunities in the world to explore my potentials and pursue my dreams. I owe my deepest gratitude to my beloved husband Abid and my three dearest children Mukhlis, Marwan and Tasneem for their infinite patience especially during my absence for many months. Their sincere flow of love has accompanied me all the way in my long struggle and has pulled me through many hurdles. Hence, this thesis is dedicated to them. ii Abstract Motivated by the belief, ‘to serve existing consumers costs less than acquiring new consumers’, firms’ marketing strategies then evolve around retaining consumers and building long-term consumer relationships. In the pursuit of acquiring consumer loyalty, enhancing consumer value has been the focus of many firms’ relationship building efforts. Hence, this study aims to understand the affect of using the Internet as a relationship marketing tool on consumer retention as well as the determinants of online consumer satisfaction affecting loyalty and retention. Although there are many factors affecting the implementation of ‘E-CRM’, that is companies’ CRM initiatives on the Internet channel; this study focuses on examining consumer perceptions towards the constituents of building online consumer relationships. Adopting a positivist approach, this research asks the following major questions: 1) How are online consumer satisfaction, loyalty and retention constructed?, and 2) How does the use of Internet technology in CRM influence the satisfaction, loyalty and retention of consumers? Data for this research were collected through questionnaire survey on Internet users in major cities of Malaysia and were analyzed using statistical techniques namely, descriptive, Structural Equation Modeling and Multivariate Analysis of Variance. The results from this study reveal that the use of Internet in building consumer relationships affects consumer satisfaction, loyalty and retention. The effectiveness of E-CRM program determines the level of which online features, such as customer service efficiency, ease of navigation, information quality, personalization and online community would be implemented on firms’ Web sites. In addition, older and welleducated users, more experienced as well as users who are involved in higher risk activities, such as online banking tend to be less tolerant. Hence, these groups of consumers seek superior quality of services from online service providers. This research contributes to knowledge in several ways. Most importantly, it demonstrates the roles of Internet technology pertinent in enhancing consumer values iii leading to long-term consumer relationships. In particular, this research highlights the critical dimensions of E-CRM program, which firms should invest in their consumer retention strategies. While repeat visits not necessarily reflect consumer loyalty and commitment to a Web site, this research advocates that when salient elements of building consumer relationships exist, service providers are more likely to improve satisfaction and gain consumer loyalty. As indicated in the E-CRM model, firms’ relationship marketing strategies should focus on identifying varying consumer expectations of service quality based on demographics, consumer level of experience with Internet technology and perceived risk. iv Table of Contents Page no. Student declaration ………………………………………………………… i Acknowledgements …………………………………………………………. ii Abstract …………………………………………………………………… . iii List of tables ………………………………………………………………… xi List of figures xvi Chapter Introduction …………………………………………………… 1.0 Introduction ……………………………………………… 1.1 Research problem …………………………………………. 1.2 Research issues and objectives ……………………………. 1.3 Justification for research ………………………………… . 1.4 Research methodology …………………………………… 12 1.5 Limitation of scope to Malaysia …………………………… 13 1.6 Conclusion and organization of thesis …………………… 14 Chapter Reviews of literature ………………………………………… . 16 2.0 Introduction ………………………………………………… 16 2.1 Internet usage and E-commerce in Malaysia ………………. 16 2.2 Current trends in Internet activities ………………………… 19 2.3 Consumers’ behaviour on the Internet: A different ………… 20 dimension from a traditional channel 2.4 Satisfaction on the Internet ………………………………… 22 2.5 Consumer loyalty on the Internet ………………………… 25 v Page no. 2.6 Consumer retention on the Internet …………………… . 28 2.7 Managing customer relationships on the Internet ………. 30 2.7.1 Customer relationship management ………………. 30 2.7.2 CRM in South-East Asia ………………………… 32 2.7.3 Electronic customer relationship management …… 38 E- CRM) 2.8 Conclusion ………………………………………………. 39 Chapter Construction of research model …………………………… . 3.0 Introduction ………………………………………………. 40 40 3.1 Causal Loop Diagram – A modeling approach ……………. 40 3.2 Theoretical framework …………………………………… 41 3.3 Development of research constructs ………………………. 45 3.3.1 Consumer satisfaction construct ……………………. 46 3.3.2 Consumer loyalty construct …………………………. 57 3.3.3 Consumer retention construct …….…………………. 62 3.3.4 E-CRM dimensions …………………………………. 71 3.4 Levels of model investigation …………………………… 76 3.4.1 Level one: Dimensions of satisfaction, retention …… 76 and loyalty 3.4.2 Level two: Causal structure of E-CRM, …………… 78 satisfaction,loyalty and retention 3.4.3 Level three: Relationships between consumer …… demographics, level of experience and perceived risk on satisfaction, loyalty and retention. vi 83 Page no. 3.4.4 Level four: Development of competing models … 88 3.5 Conclusion ……………………………………………… 91 Chapter Research methodology ……………………………………… 92 4.0 Introduction ………………………………………………. 92 4.1 Justification of paradigm and methodology ………………. 92 4.2 Survey method and administration ……………………… . 97 4.2.1 Specify the information needed ……………………. 97 4.2.2 Selection of survey method ……………………… . 97 4.2.3 Specify operational definitions …………………… 100 4.2.4 Designing the questionnaire ……………………… 102 4.2.5 105 Exploratory (pre-test) survey and revise ………… questionnaire 4.2.6 Questionnaire distribution and administration …… 107 4.3 Data analysis strategy …………………………………… . 111 4.3.1 Coding of responses ………………………………… 111 4.3.2 Cleaning and screening data ……………………… . 112 4.3.3 Selecting a data analysis strategy ………………… . 112 4.4 Ethical considerations …………………………………… 116 4.5 Conclusion ………………………………………………… 116 Chapter Data analysis ………………………………………………… 117 5.0 Introduction ……………………………………………… 117 5.1 Preliminary examination of data …………………………. 117 vii Page no. 5.1.1 Data cleaning and screening ……………………… 117 5.1.2 Descriptive analysis ……………………………… 122 5.1.3 Correlation …………………………………………. 123 5.2 Respondent profile ……………………………………… . 123 5.3 Internet usage pattern …………………………………… 125 5.4 Results from tests of research propositions ………………. 127 5.4.1 Measurement model evaluation ……………………. 127 5.4.2 Structural model evaluation ………………………… 153 5.4.3 Results from tests of competing models …………… 185 5.5 Multivariate analysis of variance and test of research ……. 193 proposition 5.6 Conclusion ………………………………………………… 201 Chapter Discussion and conclusion …………………………………… 202 6.0 Introduction ……………………………………………… 202 6.1 Discussions regarding research propositions …………… . 202 6.1.1 Dimensions of research constructs ………………… 202 6.1.2 Relationships between E-CRM and satisfaction, …… 205 loyalty and retention 6.1.3 The effect of demographics, experience level ………. 207 and perceived risk on satisfaction, loyalty and retention 6.2 Implications for theory …………………………………… 208 6.2.1 Dimensions of research construct …………………… 209 viii Page no. 6.2.2 Relationships between E-CRM and satisfaction,…… 211 and perceived risk on satisfaction, loyalty and retention 6.2.3 Demographics, experience level and perceived …… . 213 risk influence on satisfaction, loyalty and retention 6.3 Implications for practice ……………………………………. 215 6.3.1 Dimensions of satisfaction, loyalty and …………… . 215 retention 6.3.2 E-CRM influences satisfaction, loyalty and ………… 218 retention 6.5 6.4 The E-CRM model …………………………………………. 223 6.4.1 The E-CRM process …………………………………. 225 6.4.2 Market segmentation …………………………………. 227 6.4.3 Limitations in E-CRM implementation ………………. 229 Limitations and future directions of research ………………… 231 6.6 Conclusion ……………………………………………… 233 References …………………………………………………………………. 235 List of Appendices ………………………………………………………… 267 Appendix 4.1 Instrument used in semi-structured interview …… 268 with companies Appendix 4.2 Research Variables, Definition and ……………… Operationalization of Variables ix 270 Appendix 5.7 Using Structural Equation Modeling (SEM) A path diagram, first introduced by Sewall Wright is a graphical representation of hypothesized set of relationships (Cramer 2003; Tabachnick & Fidell 2001). An ellipse represents a latent (unobserved) variable while a rectangle denotes observed variable. A relationship between an indicator observe variable and a factor latent variable is indicated by an arrow from the factor to an indicator item – the construct predicts the observed variable (Byrne 2001; Cramer 2003). For example, the information quality factor predicts the level of information accuracy and relevancy. In SEM, there are two types of latent variables: exogenous and endogenous. The latter refers to latent dependent variable with at least one arrow leading to it while exogenous, or latent independent variable has no arrow pointing to it (Schumacker & Lomax 2004). In addition, a line with an arrow at both ends implies a covariance between two variables in unstandardised models, and indicates a correlation in standardized models with no hypothesized direction of effect (Tabachnick & Fidell 2001). Since no measures can be perfectly predicted, error estimates or residuals are included at both observed and latent variables respectively (Byrne 2001). The path diagrams for each of the research models are presented in Sections 5.4.1 and 5.4.2. Next, the liner models of research constructs pertaining to the analyses are discussed. Linear models of research construct Subsequent to the path diagram is the development of models indicating the relationships of constructs (and indicator variables). There are two types of linear models in SEM: measurement and structural. Measurement model. A measurement model is a model that denotes the relationship between observed variable and the construct. For example, satisfaction, retention and loyalty each are latent variables (construct) and the indicator items of each of these construct form a measurement model (Byrne 2001; Tabachnick & Fidell 2001). There are several issues imperative to the measurement model: specifying the measurement 356 Appendix 5.7 Using Structural Equation Modeling (SEM) model, number of indicators and validity tests (Hair et al. 1998). These evaluation criteria are discussed in section 5.4.5 Structural model. In turn, a structural model depicts the relationships between the constructs; E-CRM, satisfaction, retention and loyalty. A structural model relates the initial theoretical causal relationships among the latent variables developed from underlying theories (Bollen & Long 1993; Kline 1998). In this research, the structural model of latent variables was developed from theories proposed by the literature in Section 3.4. Input matrix and model estimation Input matrix. SEM uses either a variance-covariance or correlation matrix as its input data (Hair et al. 1998). Prior to inputting data, several issues need to be addressed: screened data and sample size. Data should be screened and treated for missing data, non-normality and outliers. Missing data, outliers and departure from multivariate normality pose problems to tests of causal relationships such as SEM since missing data can cause bias in the estimation process while non-normality and outliers may create bias in determining the significance of coefficient. In this study, these issues were addressed as discussed in section 5.1.1. A sufficient sample size is required to obtain a stable or meaningful parameter estimates in SEM. Researchers offer a general guideline pertaining to sample size; a sample size of less than 100 is regarded as small, medium sample size is between 100 and 200 while large sample size is more than 200 (Hair et al. 1995; Hulland, Chow & Lam 1996; Kline 1998). In this study, a sample size of 547 is appropriate to proceed with the assessment of the model. 357 Appendix 5.7 Using Structural Equation Modeling (SEM) The decision to use the variance-covariance or correlation input matrix depends on the motive of research. Variance-covariance matrix is commonly used when a study aims to test the theory applicability across different populations or samples and to validate a causal relationship (Hair et al. 1998; Hulland et al. 1996). On the other hand, a correlation matrix is appropriate when a researcher seeks to understand the pattern of relationships between constructs and make comparisons across different variables. This study has chosen variance-covariance matrix as the input matrix since it aims to examine the causal relationships of variables across a sample of respondents. Estimation technique. Sample size and normality of the data sets are important considerations in selecting an estimation technique. In this study, maximum likelihood (ML) estimation technique is preferred to other methods such as generalized least squares (GLS) and unweighted least squares (ULS). With a sample size of more than 500 and where multivariate normality is generally not violated, ML is appropriate for this study (Tabachnick & Fidell 2001); GLS requires a larger sample size (Anderson & Gerbing 1988) while ULS is sensitive to the measurement scales of observed variables (Kline 1998). Model identification After an input matrix and estimation technique have been specified, the next step is to determine whether the model is identified. Model identification refers to a definite and unique solution to each parameter of an equation model (Tabachnick & Fidell 2001). Researchers suggest that a model has to meet two requirements in order to be identified: the degrees of freedom of a model should be greater or equal to zero and that each construct should have an appropriate scale (Byrne 2001; Kline 1998; Schumacker & Lomax 1996). 358 Appendix 5.7 Using Structural Equation Modeling (SEM) Specifically, a model is said to be just-identified when it has zero degrees of freedom. In this instance, the number of parameters perfectly reproduces the sample covariance matrix, and can never be rejected (Byrne 2001) because the hypotheses about the specific paths in the model can be tested (Tabachnick & Fidell 2001). Although a just-identified model provides a perfect fit, it is not of the interest of most researchers to discover that a model lacks generalizability. Kline (1998) posits that some versions of the just-identified model are implausible due to the nature of variables. There are instances where the models in this research are just-identified and the occurrences are noted in this chapter. On the contrary, an over-identified model fits the goal for a structural model, that is, there is a positive number of degrees of freedom to ensure that the model achieve generalizability – the larger the degrees of freedom the better generalizability (Hair et al. 1998). An under-identified model produces negative degrees of freedom, that is, the model attempts to estimate more parameters despite lack of information available. In this study, none of the models are under-identified. Evaluation of model A model is evaluated on two measures: the unidimensionality and the goodness-of-fit. First, consider evaluating the measurement model. Several issues are required when evaluating a measurement model: specifying the measurement model, the number of indicators and validity tests. Specifying the measurement model concerns the act of confirming the indicators that define each construct. To so, this study conformed to Anderson and Gerbing’s (1988) guidelines and a two-step approach was performed. In the first phase, an exploratory factor analysis was conducted to assess the underlying factor structure of the scaled items (see section 4.2.5). However, a factor analysis performed in earlier steps of analysis allows each item to load on each factor, thus a factor is always a composite of all items. Subsequently, in a second phase a 359 Appendix 5.7 Using Structural Equation Modeling (SEM) confirmatory factor analysis was pursued. Confirmatory factor analysis indicators are specified to only one construct, thus a researcher has complete control over which variables define each construct. Following the theories underpinning this research, items are loaded onto each identified factor and items with poor loadings (less than 0.5) are dropped. There are instances in this study where the loadings were low and are noted as they occur (see section 5.4.1). A sufficient number of indicators should be present for a model to be identified (Hulland et al. 1996). Practically, a minimum of three or five indicators per factor should suffice (Baumgartner & Homburg 1996; Hulland et al. 1996; Tabachnick & Fidell 2001). This study adheres to this principle as much as possible, except the customer service quality construct where 11 items were used as the indicator variables to reflect the aspects of the subject matter cited in the literature (see Section 3.3.1). The extent to which indicators of each construct are correlated is an important consideration in a measurement model to ensure construct validity. Construct validity is discussed in detail in Section 4.2.6. It is concerned with two types of validity: convergent and discriminant. Following the procedures by Bollen (1989), in this study the convergent validity was evident by high loadings of items of the same construct, while discriminant validity was evident by the inter-construct correlation less than 1.0 (see Appendix 5.7a). • Unidimensionality A unidimensional model refers to a measurement model where its indicator variables load on only one factor and the measurement error terms are independent (Anderson & Gerbing 1988; Kline 1998) and reliable (Kline 1998). To measure internal reliability, Cronbach coefficient alpha is commonly used. Although alpha coefficient does not ensure unidimensonality, it is useful to assume that unidimensionalty exists. There 360 Appendix 5.7 Using Structural Equation Modeling (SEM) are some guidelines offered by the literature in relation to reliability coefficient; an alpha coefficient value of around 0.90 can be considered 'excellent', around 0.80 as 'very good,', around 0.70 as 'adequate' (Nunnaly 1978; Nunnally & Berstein 1994). Any value below 0.5 suggests that at least one-half of the observed variance may be due to random error, and measures that are unreliable should be avoided (Kline 1998). At no point in this study was the coefficient rule violated. However, coefficient alpha weights all item equally and it is not evident that a set of measures is unidimensional (Nunnally & Berstein 1994). Hence, to measure unidimensionality, it is more appropriate to use item loading standardised regression weights (Hulland et al. 1996). The standardised regression weight measures item loading on each construct (Coote 1998; Tabachnick & Fidell 2001). An absolute value of 0.70 or more is recommended, but this guideline may be readjusted to lower or higher values depending upon the research area (Hair et al. 1998; Hulland et al. 1996; Kline 1998; Tabachnick & Fidell 2001). A minimum value of 0.5 has been suggested (Tabachnick & Fidell 2001) and is deemed acceptable in this research (Hair et al. 1998). • Goodness-of-fit statistic. The next step is to describe the measures of fit used in this research. There are three types of goodness-of-fit measures: absolute fit, incremental fit and parsimonious fit measures (Hair et al. 1998). Absolute fit measures. Absolute fit measures concerns the overall model fit while ignoring the overfitting of a model that might occur. Chi-square test (χ2). One of the most basic measures of absolute fit is the likelihood ratio measured with chi-square χ2 (Hair et al., 1998). The χ2 statistic value relative to degrees of freedom is said to be significantly different from zero (p[...]... convenience (Anderson & Srinivasan 2003) For firms the increased importance of Internet channels can be seen in their contribution to disseminating information (Cho & Park 2001), enhancing consumer value (Yang & Peterson 2004), improving consumer satisfaction (Anderson & Srinivasan 2003) and retaining consumers, which in turn leads to better profitability (Reichheld & Schefter 2000) and to expanded market... Hamid & Kassim 2004) In order to have a better understanding of the roles of the Internet in enhancing consumer relationships, the links between CRM attributes delivered on the Internet (ECRM) and consumer satisfaction, loyalty and retention merit further investigation Researchers have approached this issue by examining companies’ usage of the Internet in consumer services and online communities (Adam... interactions online far exceeds any fear According to 6 An Assessment of the Internet’s Potential in Enhancing Consumer Relationships Salisbury et al (2001) and Kahneman and Tversky (1979), consumers’ associated with the interaction with an innovation, such as the Internet where the outcome is not known, perceived risks far outweighs the value of interaction in determining adoption behaviours Therefore,... Nevertheless, due to the dynamic nature of Internet technology, these constructs remain elusive and rapidly changing Failure to identify the “new” elements affecting consumer satisfaction, loyalty and retention may lead to inferior service offerings, which fall below consumers’ expectations as well as below industry standards at a point in time 2 An Assessment of the Internet’s Potential in Enhancing Consumer. .. this introductory chapter describes the research issues, objectives and research method and analysis Chapter 2 presents an extensive review of the literature pertaining to the subject matter being studied The theoretical framework underpinning this study is developed in the subsequent chapter together with eleven research propositions 14 An Assessment of the Internet’s Potential in Enhancing Consumer Relationships. .. (1999) 8 An Assessment of the Internet’s Potential in Enhancing Consumer Relationships In addition, with the further reduction in home-access broadband pricing and with the recently launched PC ownership campaign for rural areas the number of Internet users will continue to grow rapidly, suggesting good market potential for businesses that use the Internet Loh (2000) highlights that Malaysian markets... xx An Assessment of the Internet’s Potential in Enhancing Consumer Relationships CHAPTER 1: INTRODUCTION 1.0 Introduction The emergence of Internet technology, particularly the World Wide Web, as an electronic medium of commerce has brought tremendous changes in how companies compete in today’s New Economy Internet technologies provide companies with tools to adapt to changing consumers’ needs and... this study, descriptive and causal analyses were performed to find answers to the propositions of the study Exploratory study This stage involved a review of extant literature as well as discussions with experts in the subject matter The latter provided useful insights in identifying the state of importance of Internet services in the context of Malaysia and helped in gauging the potential market behaviour... point of view of both building life-time value relationships and the business cost savings involved (Peppers & Rogers 1995) The use of information technology, such as the Internet can be seen as a strategic business tool to remain competitive in the market (Sharif 2004b) The growing concern is the need for Malaysian firms to invest in core business applications, including CRM, which would boost business... describes and justifies the methodology used in this study: research design, sampling technique and the design, and administration of the survey The data analysis methods and the appropriate statistical techniques adopted are also presented in this chapter Detailed descriptions of the analysis of data are presented in chapter 5 and the findings of this research are examined, interpreted and reported Finally, . strategies then evolve around retaining consumers and building long-term consumer relationships. In the pursuit of acquiring consumer loyalty, enhancing consumer value has been the focus of many firms’. Raihan Ab Hamid, declare that the PhD thesis entitled An assessment of the Internet’s potential in enhancing consumer relationships is no more than 100,000 words in length, exclusive of tables,. School of Information Systems Faculty of Business and Law Victoria University of Technology An Assessment of the Internet’s Potential in Enhancing Consumer Relationships